Decentralized ZettaFlop Uber GPU/XPU Network for Al and Spatial Computing
summary
Announces Axelflops, a decentralized network that aims to aggregate deployed teraflop and petaflop GPUs into a ZettaFlop = 10^21 flops fabric, claiming gamers’ 100‑teraflop cards plus data‑center petaflop nodes can reach zetta scale. Describes heterogeneous realtime GPU scheduling across AMD, Intel, NVIDIA and custom XPU hardware, a dual‑token economics with dual-token architecture whitepaper forthcoming, a deep‑tech fund investing in transistors, packaging, memory and security, and a goal of <10 ms access and $70B zetta‑datacenter cost currently versus planned cheaper hardware.
AI-written summary · about
chapters (9)
- 0:00ZettaFlop vision
- 1:14Floating point’s role
- 3:56Floating point impact
- 5:24Teraflops to zetascale
- 7:11Zetascale computing inevitability
- 8:16Decentralized zetascale networks
- 9:22AxelFlops Network
- 10:49AxelFlops Network attributes
- 11:11Axel Fund and compute future
related talks
18mThe Promise of Heterogeneous Computing
23mFrom Gaming to the Metaverse
13mDemocratizing AI with Open Source Compute
12mAI-powered infrastructure:
18mThe Reality of Decentralized AI
09mAccelerating compute with software defined hardware (FPGA's)
09mDemocratizing Access to GPUs for Faster MLAI Innovation with Generative AI
22mKeynote: A64FX: Power efficient design for HPC and AI